Is RSSI a Reliable Parameter in Sensor Localization Algorithms – An Experimental Study
Ambili Thottam Parameswaran, Mohammad Iftekhar Husain, Shambhu Upadhyaya
Department of Computer Science and Engineering
State University of New York at Buffalo
201 Bell Hall, Buffalo NY- 14260, USA
{atp22, imhusain,shambhu}@cse.buffalo.edu
Abstract
Wireless sensor networks are becoming
ubiquitous and their application areas are
widening by the day. Localization algorithms
play an important role in enhancing the utility
of data collected by enabling sensors to
determine the location from which each data
packet is obtained. Localization can be done by
implementing beacon based algorithms or
signature based algorithms. Much of the
research work in this area assumes received
signal strength indicator (RSSI) as a parameter
in their localization algorithms. Since RSSI is
the key parameter, we conducted practical
experiments to assess whether RSSI could
indeed be used by localization algorithms to
determine distances between sensors. In our
experiments, we tried to calibrate and map RSSI
to distance under various conditions and
concluded that despite promising hypothetical
advantages of RSSI, even under ideal conditions
it cannot be used to determine inter nodal
distances in wireless sensor networks.
Keywords: Distance measurements, ESN,
Localization Algorithms, LQI, Motes, RSSI,
Wireless Sensor Networks
1.
INTRODUCTION
Wireless Sensor Networks (WSN) are becoming
more common in use and extensive research is
currently underway to expand the application
areas of WSN [1]. Some of the most important
application areas of sensor networks include
hurricane areas, volcano areas, forest fire areas,
etc. and these networks are usually known as
emergency sensor networks (ESN) [2]. In most
of the applications related to ESN the data
collected will become useless unless the
location information can be discerned from it.
Thus localization algorithms [3, 6] play a
significant role in the domain of wireless sensor
networks. Mainly these algorithms are
categorized into two – signature based and
beacon based. RSSI (Received Signal Strength
Indicator) [7] and LQI (Link Quality Indicator)
[4] are considered as two parameters which play
a pivotal role in the beacon based localization of
sensor nodes. Typically RSSI is a measure of
dBm, which is ten times the logarithm of the
ratio of the power (P) at the receiving end and
the reference power (Pref). Power at the
receiving end is inversely proportional to the
square of distance. Hence RSSI could
potentially be used as an indicator of the
distance at which the sending mote is located
from the receiving mote. When data from many
such neighboring motes are combined, the
location of the sending mote can be judged with
reasonable accuracy.
The literature study about RSSI shows that
RSSI is not a good candidate for localization
experiments [5]. Despite this, many papers
assume the use of RSSI in distance
measurements. Therefore, it is necessary to test
whether this hypothesis is correct or not by
creating a practical experimental setup on a
more basic level than those used in previous
studies.
The organization of this paper is as follows: The
next section presents the background
information related to RSSI and localization.
Section 3 presents experimental setup for
performing RSSI related measurements. Section
4 contains the results of our experiments.
Section 5 discusses the implications and
interprets the results we obtained. Finally, in
Section 6, we conclude this paper with plans of
future work.
2.
To put this mathematically, consider the
standard graph given in Figure 1.


BACKGROUND

RSSI and distance have a relationship which is
derived as follows.
As mentioned previously, RSSI is defined as ten
times the logarithm of the ratio of power of the
received signal and a reference power (e.g.,
1mW). i.e., RSSI  10 log P/Pref. This would
mean that RSSI  log P. It is known that power
dissipates from a point source as it moves
further out and the relationship between power
and distance is that power is inversely
proportional to the square of the distance
travelled.
In
other
words,
RSSI 
log(1/distance²).Simplifying this relationship
further we can conclude that RSSI  –log
distance. Thus if we were to plot the RSSI
measured and plot it against log of distance we
should be able to obtain an inverse linear
relationship and a graph thus generated would
serve as a standard curve that can then be
employed by a receiving mote to estimate the
distance at which a sending mote would be
located. It would also be possible to estimate the
degree of confidence with which the receiving
mote can make this estimation from the variance
at each data point. Thus in this paper we attempt
to generate a standard curve between RSSI and
sensor motes’ distances and test whether this
curve can be used to estimate the distance
between motes by randomly placing motes and
measuring the RSSI values. Such a curve if
proven successful would provide a cheap, easy
and invaluable tool to map the motes in a WSN
making the data obtained from such network
useful.

Figure 1 Standard graph showing expected relationship
between RSSI and [-log (distance)].
K is the slope of the standard plot. K can be
obtained by performing a linear regression
analysis on the data points used to generate the
standard curve. This analysis could also provide
the estimate of the constant parameter ‘A’ in the
equation that fits the data the best. Further this
linear regression fit plot can be used to estimate
the distance between two motes for a given
RSSI value based on the formula as shown
below.
RSSI = -KlogD + A
D = 10[( A− RSSI ) / K ]
3.
MATERIALS AND METHODS
We have conducted practical experiments to see
whether RSSI can be used as a candidate in
localization algorithms regardless of the
hypothetical assumptions that have been
revolving around RSSI. The aim of the
experiments was to prove or disprove that RSSI
can be used as an indicator of distance between
motes in a sensor network. The development kit
we used is Crossbow® Imote2 from Intel which
integrates 802.15.4 radio (CC2420) with a builtin 2.4GHz antenna. We have tried to create an
ideal environment to conduct these experiments
by avoiding confounding factors that would
spuriously alter RSSI. The precautions taken
include, ensuring that the surface on which the
experiments were conducted on was level and
verifying that the motes operated in full battery
power at the beginning and end of each
experiment. It was also ensured that there were
no obstacles in the communication path between
the motes causing attenuation of the signals and
that any electronic equipment that could
potentially cause interference were not present
in the vicinity of the experiment area.
The experimental set up consisted of two
crossbow
motes
programmed
with
count_send
and
count_receive
programs
written
using
visual
C#.
count_receive program was also designed
to read the RSSI value from the mote’s memory
and display it in command prompt. The motes
were arranged in a two dimensional plane and
the receiver mote is connected to a laptop
through a USB cable. The raw data was
collected and the RSSI value was extracted from
memory. We have taken 30 measurements at
each location. The mean of RSSI value obtained
at a given distance was calculated and variance
and standard error of mean were determined.
The whole process is repeated by positioning
the mote in a different location in the same
direction. The mean values were plotted against
distance and a non linear relationship was
verified. Subsequently the RSSI value was
plotted against the log of distance and linear
regression was applied to generate a standard
curve for the given motes. Further experiments
were conducted, RSSI values were obtained and
the distance was calculated from the standard
curve in order to verify its validity.
4.
RESULTS
From the data we got we have plotted a graph
between distance and RSSI. From multiple
experiments conducted in different directions
we obtained the graphs shown below.
(a) Direction - North
Figure 2 Distance versus RSSI plot showing inverse nonlinear relationship.
Figure 3 Linear regression analysis showing predicted
relationship between RSSI and
[-log (distance)] in
direction north.
Just as predicted, distance and RSSI show a non
linear relationship such that the RSSI value
decreases with increase in distance. To further
elucidate the nature of this, data was re-plotted
with –log (d). This data was then subjected to
linear regression analysis and it can be seen that
the RSSI values follow a linear relationship. It
can also be seen that as the distance increases
(negative logarithm decreases), the error in
RSSI value increases thereby decreasing its
reliability at extremes of the range of its radio
transmission radius. RSSI is measured as an
integer value and can be converted into its
corresponding dBm value by subtracting a
constant. This would mean that RSSI cannot
have decimals/fractions, hence it does not offer
enough resolution to distinguish subtle changes
in distances. But it theoretically offers
resolution to distinguish between distances that
are large enough to cause at least a unit change
in dBm power ratio at the receiving mote.
Hence we avoided using small increments of
distances in our experiments. We further looked
at RSSI in other directions to see if the behavior
seen in Figure 1 can be replicated in other
directions as well. Even though the initial result
was promising, through this multi direction
experiments it became clear that even under
ideal conditions with weather and interference
controlled, the RSSI data could not be relied
upon. At times the data showed the expected
inverse square relation with distance and at
times it didn’t.
(b) Direction – West
Figure 4 Distance versus RSSI plot showing lack of
reliability when tested in a different direction.
(c) Direction - East
Figure 5 Distance versus RSSI plot showing lack of
reliability when tested in a different direction.
The above graphs (Figure 4 and Figure 5)
exposed the problems with using RSSI in
determining the distance at which the sending
node is located.
5.
DISCUSSION
RSSI was considered as a metric in most of the
distance measurement algorithms. Even though
the ineffectiveness of RSSI is mentioned in the
literature, not many attempts were made to
implement it in a practical environment and
verity it. Elnahrawy et al have explored the idea
of using RSSI in localization algorithms
conducted in indoor environments and
determined that more complex models and
algorithms are required to improve accuracy of
RSSI based methods when used indoors [5].
However on a more basic level the reliability
and accuracy of using RSSI to determine
distance has not been extensively tested in a
simple environment (e.g., outdoors), where
interference by structures and stray wireless
signals of frequencies similar to that of the mote
being tested (2.4GHz) are not present. Such a
simplistic experimental setup would establish,
on a more fundamental level, whether RSSI can
be used for distance measurements and
localization. The initial experiments gave some
promising results and it was thought that RSSI
can be used at least to some set of WSN
application areas. The main attraction to RSSI
as a metric is that the measurement and
calculations involved with RSSI are very simple
and less complicated than other localization
metrics. But later on, after conducting a number
of experiments repeatedly in a maximum
possible ideal environment, we came to the
conclusion that RSSI cannot be used as a
reliable metric in localization algorithms
because of the following reasons:
Firstly, from the above graphs it is evident that
even in the ideal scenarios RSSI doesn’t give a
consistent behavior. So in an actual deployment
area the performance of RSSI can be much more
unreliable because of the presence of
confounding factors like interference from other
objects, attenuation caused due to barriers,
malfunctioning of the motes due to less secure
environment, power failures etc.
Secondly, it is evident from the graphs that as
the distance increase the error in measured RSSI
value increases (Figure 2). So RSSI gives very
unreliable results at the extremes of its range.
The chances for this to happen are more in an
emergency wireless sensor network because
motes can be destroyed in natural calamities,
effectively increasing the distance between two
available motes thereby making RSSI
unreliable.
6. CONCLUSION AND FUTURE WORK
Because of the reasons cited above, we come to
the conclusion that RSSI cannot be used as a
metric for distance measurements in localization
algorithms. Hence it is extremely difficult to
develop a graph between RSSI and distance
which can be used as a standard reference for
getting the distance value for a given RSSI
value.
In future we would also like to measure RSSI
with respect to multiple reference points to see
if doing so would improve localization
accuracy. Further we would like to conduct
experiments in the same environment and
measure a different parameter called LQI (Link
Quality Indicator). We plan to explore the
relationship between LQI and distance and see
if LQI would be a strong and reliable candidate
to be used in localization algorithms.
ACKNOWLEDGMENTS
This research is supported in part by U.S.
Department of Defense Grant No. H98230-081-0333.
References
[1] A.S.K. Pathan, Choong Seon Hong, and Hyung-Woo Lee.
Smartening the environment using wireless sensor networks in a
developing country. Advanced Communication Technology,
2006. ICACT 2006. The 8th International Conference, 1(2022):709-714, Feb 2006.
[2] M. Jadliwala, S. Upadhyaya, H. R. Rao, and R. Sharman.
Security and dependability issues in location estimation for
emergency sensor networks. The Fourth Workshop on eBusiness (WeB 2005), Venetian, Las Vegas, Nevada, USA,
December 2005.
[3] M. Bal, Min Liu, Weiming Shen, and H. Ghenniwa.
Localization in cooperative Wireless Sensor Networks: A
review. Computer Supported Cooperative Work in Design,
2009. CSCWD 2009. 13th International Conference on. (22-24):
438 – 443, April 2009.
[4] Sharly Joana Halder, Tae Young Choi, Jin Hyung Park, Sung
Hun Kang, Sin Woo Park, and Joon Goo Park. Enhanced
ranging using adaptive filter of ZIGBEE RSSI and LQI
measurement. iiWAS '08: Proceedings of the 10th International
Conference on Information Integration and Web-based
Applications & Services, ACM. November 2008.
[5] E. Elnahrawy, X. Li, and R. P.Martin. The limits of
localization using signal strength: a comparative study.
Proceeding of Sensor and Ad Hoc Communications and
Networks, IEEE SECON, pp. 406–414, October 2004.
[6] Chen Wang and Li Xiao. Sensor Localization under Limited
Measurement Capabilities. Network, IEEE. 3(21):16-23, MayJune 2007.
[7] Rong-Hou Wu, Yang-Han Lee, Hsien-Wei Tseng, YihGuang Jan, and Ming-Hsueh Chuang. Study of characteristics of
RSSI signal. Industrial Technology, 2008. ICIT 2008. IEEE
International Conference on, (21-24):1-3, April 2008.